Centrifuge and Numerical Investigations on Responses of Monopile-Supported Offshore Wind Turbines with Riprap Scour Protection Under Earthquakes

Abstract

Riprap scour protection is commonly employed to protect against local scour around large-diameter monopile foundations for offshore wind turbines (OWTs), and considering its influence on the static and dynamic behavior of monopiles may also provide the opportunity for further optimization of monopile design. However, only limited studies have gradually begun to investigate the contribution of scour protection to monopile bearing capacity, while its effects on the seismic responses of monopile-supported OWTs deployed in seismic zones have attracted even less attention. In this study, a series of centrifuge shaking table tests were conducted on large-diameter monopile foundations under both initial and scour protection conditions. Then, to further investigate the effects of scour protection parameters on the seismic response of offshore wind turbines, a three-dimensional finite element model was developed and validated based on experimental results. The results demonstrate that the presence of scour protection not only slightly increases the first natural frequency but also alters seismic responses of the OWT. Lower peak responses at the lumped mass are observed under Chi-Chi excitation, while lower peak bending moments of the pile occur under Kobe excitation. Additionally, seismic responses are more sensitive to variations in the scour protection length than its elastic modulus. Therefore, compared to material selection, greater emphasis should be placed on optimizing the scour protection length by comprehensively considering environmental loads, site conditions, and turbine dynamic characteristics. This study quantifies the effects of scour protection on the seismic responses of monopile-supported offshore wind turbines, which can provide new insights into seismic design optimization of offshore wind turbines with riprap scour protection.

1. Introduction

In recent years, offshore wind power has emerged as one of the most widely adopted clean energy sources. According to the newest Global Wind Report [1], total global offshore wind capacity had reached 83.2 GW by the end of 2024. As the offshore wind market develops, wind farms are being built in deeper waters and turbines with higher capacities are being deployed [2], resulting in the widespread use of large-diameter monopile foundations [3]. However, many offshore wind farms are located in coastal seismic zones, where potential earthquake activity poses a serious threat to the safe operation of wind turbines [4].

Seismic responses of monopile-supported offshore wind turbines (OWT) have been extensively investigated [4,5,6,7,8]. De Risi et al. [4] evaluated the seismic performance of typical turbines under strong seismic loading based on the finite element method. Padrón et al. [5] investigated seismic responses of 5 MW, 10 MW, and 15 MW reference turbines and found that soil-structure interaction becomes more advantageous as turbine capacity increases. In these studies, foundations were assumed to remain in their initial conditions, with the mudline at the original design state. However, scour frequently occurs around offshore wind turbine foundations due to currents and waves [9,10,11]. It reduces the pile embedment depth, which will change the static and dynamic behaviors of the turbine and affect its seismic responses. Zhu et al. [12] conducted a series of centrifuge shaking table tests to examine effects of scour holes on the dynamic characteristics and seismic responses of wind turbine foundations. Liang et al. [13] analyze the dynamic response of scoured OWTs under combined wind, wave, and seismic loading with scour considered. The results demonstrated that scour can significantly alter the dynamic responses. In practice, to mitigate the adverse effects of scour, scour protection is commonly implemented by placing rocks and stones (dumped stone riprap) around foundations as a protective measure [14]. The stone layer of defined thickness and length effectively increases the pile embedment depth. Although the stability and performance of scour protection have been extensively studied [9,15,16], its influence on the static and dynamic behavior of offshore wind turbines has received little attention.

The presence of scour protection layers has been found to influence both the bearing capacity [17,18] and the dynamic characteristics of structures [9,19], potentially affecting their seismic responses. Ma et al. [17] found that scour protection significantly reduces mudline rotation under serviceability limit state (SLS) conditions. Askarinejad et al. [18] investigated effects of scour protection on OWTs subjected to horizontal cyclic loading through centrifuge shaking table tests and finite element methods. They observed that scour protection increases soil density around the pile, enabling an approximate 10% reduction in the required embedment depth. Mayall [9] found that scour protection can enhance the stiffness and capacity of the foundation, thereby increasing its natural frequency. Similar conclusions were drawn by Ma et al. [17] through the finite element method. However, there is still a lack of comprehensive research on the combined effects of scour protection and earthquake loading on monopile-supported offshore wind turbines. Escribano and Brennan [20] investigated the stability of scour protection layers under seismic excitation using centrifuge shaking table tests. They found that the presence of scour protection reduces the likelihood of soil liquefaction, which is beneficial for foundations embedded in liquefiable soils. Nevertheless, existing seismic studies have primarily focused on the scour protection layers themselves. Consequently, research on seismic responses of large-diameter monopile-supported offshore wind turbines under scour protection conditions remains inadequate, particularly regarding the influences of different scour protection design parameters.

In this study, to investigate the effects of riprap scour protection on the seismic responses of monopile-supported offshore wind turbines, a series of centrifuge shaking table tests were conducted on the large-diameter monopile foundations of OWTs, considering both the initial and scour protection conditions, as shown in Figure 1. The experiment results are presented in Section 2. Subsequently, supplementary numerical analyses were carried out using the three-dimensional finite element method. Based on the established model, a sensitivity analysis was conducted, as presented in Section 3. Finally, the main conclusions are summarized in Section 4. This study quantifies the effects of scour protection on the seismic responses of monopile-supported offshore wind turbines, which can provide new insights into seismic design optimization of offshore wind turbines with riprap scour protection.

2. Experimental Investigation

There are several approaches for investigating the seismic responses of offshore wind turbines (OWTs), including simplified analysis methods, three-dimensional numerical simulations, and experimental techniques [21]. Compared to the first two approaches, centrifuge shaking table tests can effectively reproduce prototype stress conditions and simulate dynamic soil–structure interaction as well as realistic soil stress–strain behavior [22]. This method is considered reliable for studying the seismic responses of soil–structure systems. As a result, it has been widely applied in recent studies [6,21,22,23,24]. In this study, centrifuge shaking table tests were conducted at the Geotechnical Centrifuge Modelling Laboratory of Tongji University, with a maximum centrifugal acceleration of 50 g during dynamic testing. Meanwhile, the shaking table can generate horizontal accelerations up to 20 g within an excitation frequency range of 20 to 200 Hz. A laminated shear box was employed as the model container (500 × 450 × 550 mm), with rubber membranes lining the inner walls to effectively absorb wave reflections.

2.1. Model and Soil Properties

Table 1. Model specifications for the centrifuge tests.

Position	Property	Model	Prototype	Scale
Pile	Diameter (m)	0.08	4	1:50
	Embedded length (m)	0.36	18	1:50
	Aspect ratio	4.5	4.5	1:1
	Elastic modulus (kPa)	5.6 × 107	5.6 × 107	1:1
	Poisson’s ratio	0.3	0.3	1:1
	Flexural rigidity (kN·m2)	21.1	1.32 × 108	1:504
Superstructure	Diameter (m)	0.024	1.2	1:50
	Length (m)	0.24	12	1:50
	Flexural rigidity (kN·m2)	0.038	2.37 × 105	1:504
	Mass (kg)	0.678	8.48 × 104	1:503
	Fundamental frequency (Hz)	15.0	0.300	50:1
GroundMotion	Type	Duration(s)	Intensity(g)
Kobe	Record	30	0.05, 0.15, 0.25
Chi-Chi	Record	50	0.05, 0.15, 0.25
Acc100	Artificial	47	0.05, 0.15, 0.25

summarizes the model specifications, input ground motions for the centrifuge tests, and the similarity ratios between the model and the prototype. The large-diameter monopile foundation has a length of 22.5 m and a diameter of 4 m (in prototype scale), with an embedment ratio of 4.5. Moreover, the pile is made of aluminum alloy, with an elastic modulus of 56 GPa and a density of 2700 kg/m³. Therefore, the flexural rigidity of the model pile cross-section was 21.1 kN·m², corresponding to a prototype value of 1.32 × 10⁸ kN·m². Additionally, the model tower is made of PVC material, with a length of 0.24 m and a diameter of 0.024 m. The rotor-nacelle assembly (RNA) is simplified as a lumped mass placed at the top of the tower. Based on the mass of the prototype wind turbine, the model mass is set to 0.678 kg [25]. Subsequently, the fundamental frequency of the model turbine is determined to be 15 Hz using the transfer function, which is derived from the lumped mass and the base acceleration [6]. This corresponds to 0.3 Hz in prototype scale. Additionally, the model is embedded in dry sand, with a mean grain size (d₅₀) of 0.85 mm. Given a model pile diameter (D) of 80 mm, the resulting diameter-to-grain size ratio (D/d₅₀) exceeds 88. Therefore, the grain size effects on the soil–pile interaction can be neglected [26]. The sand is prepared through layered compaction, and the relative density is approximately 70%. Additionally, based on the measured fundamental site frequency under initial conditions, the shear wave velocity (V_s) of the soil layer was estimated to be approximately 156 m/s, as calculated using Equation (1). Further information on the soil properties is provided in [21].

$$V_s=4Hf$$

(1)

where H is the thickness of the sand layer, and f is the fundamental site frequency.

2.2. Scour Protection Design

To investigate the effect of scour protection on seismic responses of the large-diameter monopile foundation for OWTs, two test scenarios were designed, as shown in Figure 1a,b. The initial condition refers to the case where the soil remains in its original design state and is not affected by scour or scour protection. In contrast, the scour protection condition involves a protection layer consisting of rocks and stones around the foundation. Dumped stone riprap is a widely adopted scour protection method for hydraulic structures such as bridges and offshore wind turbines, owing to its simple construction, easy material sourcing, low cost, and adaptability [14]. As shown in Figure 1c, the scour protection layer consists of a primary armor layer and an underlying filter layer. The armor layer provides scour protection stability and resists deformation induced by hydrodynamic forces, thus requiring relatively large median particle sizes for sufficient weight. In contrast, the filter layer is placed to maintain interface stability [27]. When designing a scour protection layer, it is necessary to specify the median particle size (d₅₀), thickness, and length of both the armor and filter layers. Specifically, the design of the armor layer is governed by wave and current conditions, while the filter layer design is based on the grain sizes of the armor layer and the underlying sand [20]. Previous scour protection cases for offshore wind turbine foundations indicate that the median grain size of the armor layer (d_50,a) generally ranges from 0.15 to 0.425 m [14]. In this study, gravel with a median grain size of 7 mm (prototype scale: 0.35 m, within this range) was selected for the armor layer, and sand with a median grain size of 1.5 mm was chosen for the filter layer, resulting in a d_50,a/d_50,f ratio of 5 [20]. The grain size distribution curves are presented in Figure 2. According to DNV standards and prior studies [15,18,28], the protection length was set to 5D (D means the diameter of monopile), while the armor layer thickness H_a was set to 3d_50,a and the filter layer thickness H_f to H_a/2 [14,29].

2.3. Seismic Motions

Three seismic motions with different spectral characteristics were selected in the tests: Kobe, Chi-Chi, and Acc100 motions. Specifically, Kocaeli and Kobe motions are existing seismic records derived from significant earthquakes that occurred in Kocaeli, Turkey [30], and Kobe, Japan [31], respectively. These motions are based on the ground motion recordings from the actual seismic events that took place in these regions. Conversely, the Acc100 motion is an artificial motion calculated based on the geological conditions of the southeastern coastal regions in China [21]. To assess the influence of input intensity, three seismic motions were scaled to target peak ground accelerations of 0.05 g, 0.15 g, and 0.25 g using a waveform reproduction system, corresponding to 2.5 g, 7.5 g, and 12.5 g on the shaking table under 50 g of centrifugal acceleration. Considering the operational frequency range of the shaking table (20–200 Hz), three seismic inputs were first scaled in time and amplitude in accordance with similitude principles, and band-pass filtering was applied to eliminate frequency components outside the effective range of the shaking table. The final input motions used in the tests are shown in Figure 3 (in prototype scale). Acceleration time histories and their corresponding Fourier spectra are presented in Figure 2 (in prototype scale). It can be seen that the Kobe motion contains more low-frequency components, whereas the Chi-Chi and Acc100 motions are characterized by higher predominant frequencies. The durations of the Kobe, Chi-Chi, and Acc100 motions are 30 s, 50 s, and 47 s, respectively. Subsequently, identical excitation sequences were applied to both scenarios depicted in Figure 1 (initial condition and scour protection condition), following an ascending intensity order of Kobe, Chi-Chi, and Acc100 motions. Before the seismic inputs, a white noise sweep with an amplitude of 0.01 g was performed to obtain the frequency characteristics of the site and the monopile-supported offshore wind turbine.

2.4. Measured Results

2.4.1. System Frequency

The fundamental frequency of offshore wind turbines is a key parameter in wind turbine design, which can reflect their dynamic characteristics to a certain extent. In this study, the fundamental frequency of the offshore wind turbine is determined based on the transfer function, which is derived from the measured accelerations [6,21]. Figure 3a presents the transfer function from the lumped mass to the base of the shaking table, along with its corresponding smoothed curve, where the peak frequency represents the first natural frequency of the offshore wind turbine. Comparison between the initial and scour protection conditions shows an increase in the first natural frequency from 0.300 Hz to 0.305 Hz, corresponding to a 1.7% rise. It indicates that the offshore wind turbine with the scour protection layer exhibits “stiffer” qualities, as the presence of the layer both increases the embedded depth of the pile and the pile–soil lateral stiffness. Figure 3b shows the transfer functions from the soil surface to the base under both conditions. It can be observed that the site frequency increases by 1.8% with the presence of the scour protection layer compared to the initial condition. This further confirms that the presence of the scour protection layer enhances foundation stiffness, thereby increasing the first natural frequency of the offshore wind turbine.

2.4.2. Site Responses

Figure 4 illustrates the peak acceleration responses at the ground surface (measured by A4) under various seismic excitations for the initial condition. The gray dashed line represents a reference with a slope of 1 and passing through the origin. Overall, under all three excitation levels (0.05 g, 0.15 g, and 0.25 g) of the Kobe, Chi-Chi, and Acc100 motions, the peak ground accelerations (PGA) lie above the reference line, which indicates a pronounced amplification effect at the site. Specifically, the Acc100 excitation results in the most significant amplification, while the amplification under Kobe excitation is comparatively weaker. Figure 5 presents the site responses under the Chi-Chi 0.25 g excitation for the initial condition, including acceleration time histories and corresponding Fourier spectra. As shown in Figure 5a, the peak acceleration at the surface is 72% higher than that at the base, which indicates a clear amplification effect from base to surface. Additionally, Figure 5b shows that as the soil thickness decreases, the predominant frequency shifts from approximately 3 Hz to the range of 1–2 Hz, demonstrating a filtering effect of the soil. Given that the site frequency is approximately 1.9 Hz (shown in Figure 3b), spectral components of input motions near this frequency are significantly amplified.

2.4.3. Structure and Foundation Responses

Figure 6 illustrates the peak acceleration and displacement responses at the lumped mass under all seismic excitations. As shown in Figure 6a, the peak acceleration response under Chi-Chi excitation is the lowest compared to Kobe and Acc100 motions. It may be attributed to the fact that Chi-Chi motion contains richer high-frequency components relative to Kobe and Acc100, as shown in Figure 6b. Meanwhile, the fundamental frequency of the offshore wind turbine is approximately 0.3 Hz, which aligns more closely with the dominant frequencies of Kobe and Acc100 motions than with those of the Chi-Chi motion. A similar trend is observed in Figure 6b, where the displacement response at the lumped mass is lowest under the Chi-Chi excitation. This is because displacement reflects the cumulative effect of acceleration. In the frequency domain, integration reduces high-frequency components and enhances low-frequency ones. Figure 7 illustrates the envelope curves of pile bending moment responses under different amplitude excitations. The maximum bending moment consistently occurs at approximately 0.62 times the embedded depth of the monopile. As shown in Figure 2b and Figure 3b, the dominant frequency of the Kobe motion is significantly farther from the site’s fundamental frequency than those of the Chi-Chi and Acc100 motions. Consequently, the pile bending moment response under the Kobe excitation exhibits the lowest peak. It is worth noting that the existence of the scour protection layer leads to an increasing trend in the peak pile bending moments. For instance, under Kobe 0.05 g, Chi-Chi 0.05 g, and Acc100 0.15 g excitations, the maximum pile bending moments increase by 12.9%, 12.1%, and 19.6%, respectively. It is likely attributable to the enhanced soil stiffness provided by the scour protection layer, which intensifies the soil reaction forces on the monopile under the same displacement and thereby generates greater internal forces within the pile.

3. Numerical Analysis

3.1. Finite Element Method and Validation

The finite element method can obtain more accurate results based on refined three-dimensional numerical models and suitable constitutive models (of the relationship between stress and strain). Therefore, it is widely adopted in the seismic analysis of offshore wind turbines [32,33,34]. To further investigate the internal force and deformation characteristics of the monopile foundation with scour protection under seismic excitation, a supplementary numerical analysis was conducted using the three-dimensional finite element method based on the ABAQUS platform. Meanwhile, the implicit dynamic algorithm was adopted, and its detailed formulation can be found in the ABAQUS user manual [35]. Before the numerical analysis, the methodology was validated by comparing numerical calculations with the experimental results of centrifuge shaking table tests. The numerical model was developed as an experimental model to replicate the centrifuge test conditions. It strictly followed the prototype dimensions corresponding to the experimental setup, with a monopile diameter of 4 m and an embedment ratio of 4.5, as listed in the “Prototype” column of Table 1. Solid elements were employed to model the soil, monopile, superstructure, and scour protection layer [17]. The element type is specified as eight-node linear brick elements (C3D8R), which utilize reduced integration techniques to achieve a good balance between efficiency and accuracy, and are especially appropriate for simulating nonlinear material behavior and large deformation problems [34]. Moreover, the ideal elastoplastic Mohr–Coulomb model was employed to simulate the soil and scour protection layer. This constitutive model is extensively adopted in practical geotechnical engineering and is regarded as one of the preferred models in industry because of its simplicity and satisfactory accuracy [32]. The shear modulus G_s of the sand was calculated based on the measured shear wave velocity V_s, as shown in Equation (2). Meanwhile, the elastic constitutive model was adopted to simulate the monopile foundation and superstructure.

Table 2. Model specifications and earthquakes for the numerical simulation.

Position	Geometry	Value (m)	Material Property	Value
Monopile	Outer diameter	4	Constitutive model	Linear elastic
	Inner diameter	3.8	Elastic modulus (kPa)	5.6 × 107
	Length	22.5	Poisson’s ratio	0.33
	Embedded length	18	Density (kg/m3)	2.7 × 103
Tower	Outer diameter	1.2	Constitutive model	Linear elastic
	Inner diameter	0.8	Elastic modulus (kPa)	3.3 × 106
	Height	12	Poisson’s ratio	0.33
			Density (kg/m3)	1.4 × 103
Mass	Edge length	2.25	Constitutive model	Linear elastic
			Mass (t)	84.8
Soil Layer	Length	25	Constitutive model	Mohr–Coulomb
	Width	22.5	Elastic modulus (kPa)	1.1 × 105
	Total thickness	20.5	Poisson’s ratio	0.3
	Layer thickness	1.2	Density (kg/m3)	1.8 × 103
			Internal friction angle (°)	35
GroundMotion	Type	Duration(s)	Intensity(g)
Kobe	Record	30	0.05, 0.15, 0.25
Chi-Chi	Record	50	0.05, 0.15, 0.25
Acc100	Artificial	47	0.05, 0.15, 0.25

provides detailed information on the numerical model. Additionally, surface-to-surface contact was adopted to model the pile–soil interface. The normal contact behavior was defined as “hard contact,” which allows separation when tensile stresses occur at the interface elements. The tangential contact behavior was simulated using the penalty contact method based on Coulomb’s friction theory. Regarding the boundary conditions, displacements in all three directions (x, y, and z) at the model base were constrained, and the constraint in the x-direction (aligned with the seismic acceleration) was released during the seismic loading step. To reduce the influence of reflected waves at the vertical boundaries during seismic excitation, multi-point constraint (MPC) boundary conditions were implemented to accurately simulate the free-field soil response [33].

$$G_s={\rho }_s\cdot V_s^2$$

(2)

where ρ_s is the soil density, and V_s is the shear wave velocity obtained from the centrifuge model tests.

Table 3. Comparison of experimental and numerical fundamental frequencies.

Method	Fundamental Frequency (Hz)	Relative Error
Numerical	0.304	—
Experimental	0.300	1.3%

presents a comparison of the first natural frequency of the monopile-supported offshore wind turbine between numerical and experimental results under the initial condition. The discrepancy is within 1.3%, indicating that the developed numerical model can accurately capture the dynamic characteristics of the prototype offshore wind turbine system. Additionally, the model was subjected to seismic loading using the measured base acceleration from the centrifuge tests, which corresponds to the actual input motion applied to the shaking table, as shown in Figure 3. Comparisons between the numerical and experimental acceleration responses were conducted for both the soil and the structure. Due to space limitations, Figure 8 illustrates only the accelerations under the Kobe excitation, including both the initial and scour protection conditions. Acceleration time histories show good agreement at various locations in the soil, pile, and superstructure, demonstrating the numerical model’s reliability in reproducing seismic wave propagation. Figure 9 shows the comparison of pile bending moment envelopes. The peak bending moments obtained from both methods are similar in magnitude, though the numerical model predicts the maximum moment at a slightly higher position than that observed in the centrifuge tests. This discrepancy may be attributed to the simplified assumption of uniform shear modulus in the numerical model, whereas soil stiffness increases with depth in reality. Nevertheless, the deviation is minor. Overall, the developed numerical model not only reflects the dynamic characteristics of the offshore wind turbine but also reasonably reproduces its seismic responses. Therefore, the reliability of the 3D finite element model developed on the ABAQUS platform for seismic analysis has been demonstrated.

3.2. Discussion with Numerical Analysis

Figure 10a,c show the maximal principal logarithmic strain distribution within the monopile domain at the time of maximum lateral displacement at the nacelle position under seismic excitations of Kobe and Chi-Chi motions, respectively. Figure 10b,d present the Mises stress distribution within the monopile domain at the same time. It is evident that both strain and stress under the Kobe excitation are greater than those induced by the Chi-Chi motion. This is attributed to the natural frequency of the offshore wind turbine being closer to the dominant frequency of Kobe motion, resulting in resonance amplification of seismic responses. In contrast, the Chi-Chi motion contains richer high-frequency components, leading to comparatively lower seismic responses. Additionally, the presence of the scour protection layer changes the stress and strain distribution within the monopile foundation. Meanwhile, the characteristics of input seismic motions also significantly influence the internal force distribution, particularly in terms of amplitude and frequency content.

Figure 11a,c illustrate the distribution of maximal principal logarithmic strain in the soil domain at the time of maximum lateral displacement at the nacelle under Kobe and Chi-Chi excitations, respectively. Figure 11b,d show the corresponding Mises stress distributions. Compared to the Chi-Chi motion, the soil strain under Kobe excitation is lower, while the stress levels remain similar. It indicates that soil strain is more sensitive to variations in the frequency content of seismic motions than soil stress level. In addition, the presence of the scour protection layer significantly reduces soil strain near the mudline and slightly increases stress in the shallow layers under both high- and low-frequency seismic excitations. It can be attributed to the constraint imposed by the scour protection layer on the surface soil, which reduces its deformability. Furthermore, the existence of the scour protection layer increases the vertical effective stress in soil and enhances the horizontal stiffness of the soil–pile system, thereby causing a slight increase in soil stress.

3.3. Parametric Analysis

To evaluate the relative importance of scour protection layer design parameters on the seismic responses of offshore wind turbine large-diameter monopile foundations, a sensitivity analysis was conducted based on the validated finite element model described above. Effects of scour protection properties (elastic modulus), thickness, and length were discussed in this study, and all numerical test conditions are summarized in

Table 4. Summary of scour protection conditions.

Condition ID	Elastic Modulus (MPa)	Thickness (m)	Length (D)
T0	150	1.05	5
E1	100	1.05	5
E2	200	1.05	5
H1	150	0.55	5
H2	150	1.55	5
R1	150	1.05	3
R2	150	1.05	4

. Seismic excitations were applied using Kobe, Chi-Chi, and Acc100 motions, all with a peak acceleration of 0.15 g. The resulting acceleration and bending moment responses were subsequently analyzed for comparative evaluation.

3.3.1. Effect of Scour Protection Properties

Figure 12 presents the response envelopes of offshore wind turbine acceleration and the monopile bending moment with scour protection layers of varying elastic moduli under Kobe, Chi-Chi, and Acc100 seismic excitations, respectively. As shown in Figure 12a–c, peak acceleration decreases as the elastic modulus increases, while the peak location remains nearly unchanged, consistently occurring near the mid-height of the tower. Specifically, when the elastic modulus of the scour protection increased by 100%, the peak acceleration decreased by 0.8%, 0.4%, and 3.8% under Kobe, Chi-Chi, and Acc100 excitations respectively, indicating minimal variation. Furthermore, as shown in Figure 12d–f, the peak bending moment increased by 1.8%, 3.4%, and 2.0% under Kobe, Chi-Chi, and Acc100 excitations respectively. There are also minimal changes and the envelope shapes remained largely consistent. In summary, acceleration and bending moment responses of the offshore wind turbine exhibit low sensitivity to variations in the elastic modulus of the scour protection layer under seismic excitation. Consequently, in practical engineering, the selection of scour protection materials can be primarily guided by hydrodynamic conditions and scour parameters, and there is no need to place undue emphasis on the impact of their elastic modulus on seismic responses of offshore wind turbines.

3.3.2. Effect of Scour Protection Thickness

Figure 13 illustrates the response envelopes of offshore wind turbine acceleration and the monopile bending moment with scour protection layers of varying thicknesses under Kobe, Chi-Chi, and Acc100 seismic excitations, respectively. As shown in Figure 13a–c, when the thickness of the scour protection increased by 181%, the peak acceleration decreased by 1.9%, 7.5%, and 9.2% under Kobe, Chi-Chi, and Acc100 excitations, respectively. Compared to changes in the elastic modulus, variations in the thickness of the scour protection layer have a more pronounced effect on acceleration responses of OWTs, particularly under the Acc100 seismic excitation. Furthermore, Figure 13d–f indicate that with a 181% increase in scour protection thickness, the peak bending moment of the monopile foundation increased by 3.5%, 5.9%, and 10.8% under Kobe, Chi-Chi, and Acc100 excitations, respectively. Meanwhile, it is also observed that under the same rate of increase, the increment of the peak bending moment of the monopile is more significant when the scour protection thickness increases from 0.55 m to 1.05 m. It indicates that further increases have a diminishing effect on the bending moment responses of the monopile foundation when beyond a certain thickness of the scour protection layer. Additionally, compared to the bending moment along the pile in deeper soil layers, changes in the scour protection thickness have a greater impact on the bending moment in shallow soil layers than in deeper layers along the pile. The reason is that the presence of the scour protection more significantly affects the stress distribution in the shallow soil, as illustrated in Figure 11b,d.

3.3.3. Effect of Scour Protection Length

Figure 14 shows the response envelopes of offshore wind turbine acceleration and the monopile bending moment with scour protection layers of varying lengths under Kobe, Chi-Chi, and Acc100 seismic excitations, respectively. As shown in Figure 14a–c, when the scour protection length increases by 67%, the peak acceleration decreases by 9.6% and 5.3% under Chi-Chi and Acc100 excitations, respectively, while the peak acceleration increases by 1.2% under Kobe excitation. This phenomenon indicates that different seismic spectral characteristics also significantly affect the seismic responses of offshore wind turbines. Furthermore, Figure 14d–f reveal that with increasing protection layer length (67%), the peak bending moment increases by 5.0%, 5.4%, and 14.8% under Kobe, Chi-Chi, and Acc100 excitations, respectively. Consistent with the findings in Figure 13, changes in scour protection length have a greater impact on the bending moment at shallow soil layers than at deeper layers along the pile. Overall, compared to variations in the elastic modulus and thickness of the scour protection layer, changes in its length significantly affect the acceleration and bending moment responses of offshore wind turbines under seismic loading. However, to evaluate the development of this effect, it is necessary to comprehensively consider the fundamental frequency of the offshore wind turbine and the dynamic characteristics of the seismic motions, rather than relying solely on scour protection design parameters.

4. Conclusions

Monopile-supported offshore wind turbines (OWTs) are widely installed in seismic regions, and large-diameter monopiles are increasingly adopted with growing wind turbine capacity. Additionally, scour phenomena commonly occur around pile foundations due to water-current actions, and scour protection consisting of rocks and stones is typically placed around the foundation to ensure the stability of wind turbines. Considering the influence of scour protection on the static and dynamic behavior of monopiles may also provide the opportunity for further optimization of monopile design. However, only limited studies have begun to investigate the contribution of scour protection to monopile bearing capacity, while its effects on the seismic responses of monopile-supported OWTs deployed in seismic zones have received even less attention. In this study, a series of centrifuge shaking table tests were conducted on large-diameter monopile foundations under both initial and scour protection conditions. Particular attention was given to system frequencies, site responses, and structure responses. To further investigate the effects of scour protection parameters on seismic responses of offshore wind turbines, a three-dimensional finite element model was developed and validated based on experimental results. Based on a series of centrifuge tests and numerical analyses, the following conclusions are as follows:

• Centrifuge test results indicate that the presence of the scour protection slightly increases the first natural frequency of the OWT due to enhancing the lateral stiffness of the soil–pile system, thereby altering the dynamic characteristics of the structure. Moreover, the soil exhibits pronounced amplification and filtering effects under the seismic excitations considered in this study.
• Lower peak responses at the lumped mass are observed under Chi-Chi excitation, likely due to its dominant high-frequency components being further from the fundamental frequency of the OWT. In contrast, lower peak bending moments of the pile are observed under Kobe excitation, as its predominant frequency deviates more from the site frequency. This highlights the importance of frequency-domain compatibility between input motion and system response. Moreover, the presence of the scour protection leads to a 10–20% increase in peak bending moments, suggesting that increased stiffness at the pile–soil system results in larger inertial force to the foundation.
• Based on the validated 3D finite element model, stress and strain distributions of the OWT and soil under seismic loading were investigated. Results indicate that soil strain is more sensitive than soil stress to changes in seismic frequency content, and the presence of scour protection reduces soil strain near the mudline while slightly increasing stress in the shallow soil. This is mainly due to the constraint imposed by the scour protection layer, which reduces surface soil deformability and increases vertical effective stress.
• In this study, the peak acceleration of the large-diameter monopile-supported OWT occurs at the tower, while the maximum pile bending moment appears near 0.62 L (L, embedded depth). Parameter analysis indicates that seismic responses are more sensitive to variations in the scour protection length than its elastic modulus, and different seismic excitations also affect this behavior. This is because variations in protection length affect the extent of lateral confinement and energy dissipation along the pile. Overall, material selection of the scour protection has limited influence on seismic responses of OWTs. Greater emphasis should be placed on optimizing the length of the scour protection layer, taking into account environmental loads, site conditions, and the dynamic characteristics of the OWT.

This study quantifies the effects of scour protection on the seismic responses of monopile-supported offshore wind turbines, which can provide new insights into seismic design optimization of offshore wind turbines with riprap scour protection. Nonetheless, this study has certain limitations that should be recognized. Firstly, the experiments were conducted using dry sand, whereas offshore monopiles are typically embedded in saturated soils. The effects of soil saturation and potential liquefaction on both the foundation and the scour protection layer were not considered. Additionally, the numerical model does not account for particle interactions within the scour protection layer. Instead, it adopts a Mohr–Coulomb constitutive model following previous studies, which treats the layer as a continuous medium. Future work will incorporate saturated soil conditions and develop more advanced numerical models to better capture seismic responses of monopile-supported offshore wind turbines with riprap scour protection.